摘要
本文建立关于分布自由的、带滑动系数的逐步回归集成最优方程具有如下特点:(1)对数据进行秩变换,从而排除正态、线性模型的限制;(2)利用“秩和”作为新因子,集中了几个因子的信息,从而利用了大量因子的信息;(3)在不同置信水平下建立多个秩数逐步回归方程,然后综合各个逐步回归方程的信息,建立分布自由的动态量优集成预报方程,其集成系数与各时刻的预报值有关,因而能利用各个时刻观测值的信息,提高预报的可靠性.最后给出一个实例.
A stepwise regressive integrated optimal prediction method with freedom of distribution and slide coefficient is presented in this paper.The method is characterized by 1)transforming data into rank so as to remove the confinement of normal linear model; 2) poolinformation from several factors by “sum of rank”.So as to make use of infomation from a last amout of factors; and 3) forming multi- pie stepwise regressive equations with cyclomatic numbers under different confidence levels,and then forming a freely distributed dynamic optimal integrated prediction method by synthesizing the informa- tion from these stepwise regressive equations.Its integration coefficient relates to the predicted values at different time.thus the reliability can be promoted by making use of information of observed values at different time.The method is exemplified at the end of the paper.
出处
《华侨大学学报（自然科学版）》
CAS
1992年第1期26-33,共8页
Journal of Huaqiao University(Natural Science)
关键词
滑动系数
秩和因子
预报方程
slide coefficient
factor of rank-sun
prediction equation
